the peak period of risk for developing AD. The chromosomal region identified in the current association analysis partially overlaps with the alcohol dependence linkage signal on chromosome 3 previously reported in a COGA dataset (Foroud et al., 2000b) from which these subjects were drawn. In contrast to the GWAS dataset used in this study the linkage sample consisted of both non-Hispanic white families and African American (AA) families. Indeed. The linkage signal is apparent in each population when analyzed separately suggesting that there is an AD risk locus in this region in both populations. However, the imputed and genotyped SNPs showing the strongest association in EAs exhibit much lower LD (D’ = 0.84, r2 = 0.12, 1000 genome pilot dataset) with each other in subjects of African ancestry. As a result, although we have linkage data and exome chip data in the AA families we cannot accurate impute the associated SNPs. Furthermore, assuming these SNPs tag the functional variation but are not the functional alleles the differences in LD may mean that these SNPs are not in LD with the functional alleles in AAs. Further genotyping and sequencing in the AA families will be needed to determine whether the linkage
